Co-optimized elasto-geometrical and hand-eye calibration of industrial robots with integrated dual laser profile scanners

Hand-eye calibration is a fundamental prerequisite for vision-based applications in industrial robotics. While this issue is largely addressed for 3D cameras, it remains a challenge for Laser Profile Scanners (LPSs) due to their inherent 2D measurement limitations. Performing hand-eye calibration with an uncalibrated robot often results in inaccurate calibration outcomes. Traditional methods for calibrating the robot prior to hand-eye calibration using highly accurate optical systems are often prohibitively expensive and time-consuming, making them impractical for the high demands of modern production environments. This study proposes a novel methodology for co-optimized elasto-geometrical calibration of the robot, while modeling joint compliance, and the hand-eye calibration of a dual LPS mounted on the robot’s end-effector. The two LPSs are configured such that their laser lines intersect on the object, thereby forming an intersecting Laser Line System (LLS). This necessitates the calibration of the LPSs relative to each other, which constitutes eye-to-eye calibration in the context of multiple LPSs. Therefore, this work also introduces an approach for the eye-to-eye calibration of the LPSs. Both proposed approaches leverage the iterative closest point (ICP) algorithm within a bilevel optimization framework and utilize an artifact specifically designed for the fast and stable calibration of LPSs. Notably, these methods are extensible to multiple laser lines with diverse configurations and are applicable to analogous calibration artifacts. To the best of the authors’ knowledge, this is the first study to address the eye-to-eye calibration of LPSs and to explicitly incorporate joint stiffness effects into the hand-eye calibration of LPSs. The experimental results demonstrate that the proposed calibration strategy achieves a registration error of 0.154 mm when using a 3D-printed artifact, while employing a high-precision artifact results in an improved accuracy with a registration error as low as 0.068 mm.